Miniaturization of Hvdrowocessina Catalvst Testing System -Sie-1996-AIChE_Journal

8/18/2019 Miniaturization of Hvdrowocessina Catalvst Testing System -Sie-1996-AIChE_Journal

1/10

Min ia tu r i za t ion

of

H v d r o w o c e s s i n a C at alv s t

Tes t ing Sys tem s : Theory and Prac t i ce

S.

T.

Sie

Faculty of Chemical Technology an d M aterials Science, Delft Un iversity of Technology,

2628 BL Delft, T he N etherlands

Main factors limiting downscaling of fi ed -b ed integral reactors are discussed, which

operate with a gas stream

or

with gas and liquid i n trickle flow, as used in hydroproc-

essing

of

oil. Criteria developed for a sufjiciently close approach to plug flo w and

for

good contacting

of

the catalyst show th at they can be m et i n very small reactors, both

for processes with reactants in the gas phase and for trickle-flow processes, provided

that the catalyst bed is diluted with fine inert material in the latter case. Experimental

tests show that microreactors with typical catalyst volumes

of 5-10 mL

can be used to

obtain representative results that can very well match data from industrial reactors. It

appears to be feasible to m iniaturize catalytic test reactors jkrthe r to “nanosca1e”reac-

tors with as little as 0.2-0.4 mL

of

catalyst while maintaining th e results to be me an-

ingful. Even though there are advantages as well as limitations, miniaturization can

enhanc e safety and reduce manpower.

Introduction

Miniaturization of equipment has many advantages and can

in some cases even revolutionize fields of technology. Ekam-

ples of drama tic impacts of miniaturization are in the field of

biochemistry where the advent

of

microscale chromato-

graphic separations paper, thin layer, and capillary high-per-

formance liquid chromatography HPLC)) and other mi-

croscale analytical techniques have been crucial to progress

in this field for the last decades, a nd in the field of electronic

computing, where miniaturization is causing a revolution in

information technology.

In the industrial chemical laboratory, too, reducing the

scale of experimentation without detracting from the value of

the results can be quite important, although the conse-

quences of miniaturization may not be as spectacular as in

the cases just cited. The incentives behind downscaling are

manifold, and may include the ones below:

Che aper equipment to construct and install

Fewer materials to consume, store, and dispose of

Fewer demands

on

laboratory infrastructure, including

Intrinsically safer

May promote better, more precise, working practices.

The present article deals with the topic

of

scale reduction

in research on catalytic processes. Catalysis, being a molecu-

lower utility requirements

lar phenomenon, can in principle be studied on a very small

scale. For example, a microgram sample of a compound hav-

ing a molecular weight of 600 consists of as many as about

10’’ molecules, which is more than sufficient for this sample

to be truly representative.

As

a matter

of

fact, fundamental

catalysis studies on single-crystal surfaces equivalent to one

microgram of a practical catalyst often involve minute

amounts of reactants, and these studies are made possible by

the high sensitivity of modern analytical techniques such as

mass spectrometry, gas chromatography, and spectroscopic

techniques that are also based o n molecular phenom ena.

In contrast to these fundamental studies, research and de-

velopment studies on industrial catalytic processes are not

only concerned with understanding the basic chemistry of

these processes but also with relating the information gener-

ated in the laboratory with industrial practice. Since the com -

position of practical feedstocks such as petroleum fractions

may be very complex, and in view of the complicated reaction

networks involved, a fundamental kinetic approach is only

possible in a limited number of cases, notwithstanding the

possibilities offered by modern analytical techniques and

powerful computers. Therefore, there is a need for the ability

to simulate the performance of an industrial reactor in the

laboratory,

3498 December 1996

Vol.

42, No. 12 AIChE Journal


2/10

Traditionally, the laboratory equivalent of an industrial re-

actor as used in hydroprocesses meters in width and tens of

meters in height) was a pilot plant of much smaller dimen-

sions but still very large according to laboratory standards

e.g., a catalyst bed a few centimeters wide and several me-

ters high). The subject of the present article is the downscal-

ing of this pilot plant to more convenient dimensions while

retaining the validity of the results obtained. The article will

focus on the downsizing of fixed-bed reactors, which are

among th e types of reactors m ost widely used in the pe troleum

industry. Fixed-bed reactors operated with a gaseous stream

of reactant as well as with a com bination of gas and liquid,

namely, trickle-flow reactors, will b e c onsidered.

Downscaling Approach for Continuous-Flow

Fixed-Bed Reactors

For a fixed-bed reactor operated under steady conditions,

the important scaling factors are the dimensions of the cata-

lyst bed and the catalyst particles. To simulate an industrial

reactor generally operated as an integral reactor achieving

practical levels of conversion of the reactant, the laboratory

reactor must also be operated as an integral reactor giving

the same conversion. The approach generally taken is to op-

erate th e laboratory reactor at the same space velocity as the

industrial one. Since at a constant space velocity the linear

velocity of the reactant stream is directly proportional to the

bed length, a reduction of bed length will reduce the linear

velocity and th e Reynolds numb er accordingly

Re=

d p

*

u/q,

in which d p is the particle diameter, the superficial veloc-

ity, and

q

the dynamic viscosity). In principle, this means that

the hydrodynamics in the commercial reactor and the shorter

laboratory reactor will be different.

Hydrodynam ic factors play a crucial role in tha t they affect

the reaction rates in an interactive way with the intrinsic

chemical kinetics, so one cannot reduce the reactor length,

or

at least not by a large factor, without losing representative-

ness of the laboratory reactor. The only dimension that can

be reduced is the reactor width, down to a size where the

wall effect begins to affect results to be discussed later).

These considerations are m ore o r less implied in th e sizing of

the traditional pilot plant for fixed-bed processes: reactor

lengths of 5 to 10 m and reactor diameters of 4 o 10 cm.

Fortunately, however, in most fixed-bed processes of prac-

tical interest, in particular th e widely applied fixed-bed proc-

esses in the petroleum industry, hydrodynamics play a less

important role than chemical kinetics and intraparticle diffu-

sion in determining conversions, In the general situation of

slow or moderately fast reactions occurring over catalyst par -

ticles with effectiveness factors not much below one, the dif-

fusion paths within the catalyst particle are of the order of

the particle radius, while the effective film thicknesses out-

side the particle involved in gas-solid, liquid-solid, or

gas-liquid mass transfer will b e substantially sm aller. This

point is illustrated in Table 1, which gives the relative magni-

tude of the various mass-transfer resistances in a typical case

of trickle-flow hydrotreating. Un der these circumstances, hy-

drodynamics are only

of

importance insofar as they deter-

mine pressure drops, holdups in multiphase flow, and in cer-

tain cases distribution of fluids over the reactor cross section

and heat-transport phenomena.

Table

1.

Relative Mass-Transfer Resistances Estimated for a

Typical Case

of

Gas Oil Hydro desulfurization n Trickle

Flow

Resistance of Total

Negligible

2

21

From bulk gas to

G/L

interface

From G/L interface to bulk liquid

From bulk liquid to ext. cat. surface

Intraparticle diffusion and reaction 77

When the desired function of the laboratory reactor does

not include being representative o n the last-mentione d points,

but is restricted to representativeness in a chemical reaction

sense only, considerable f urth er dow nscaling is possible.

As

a

starting point, one may assume that a well-designed indus-

trial fixed-bed reactor is a close approximation of an ideal

integral rea ctor tha t fulfills the following criteria:

1.

All volume elements of the feed must spend the same

time in the reactor so as to contribute equally to the overall

conversion. In other words, the reactor should be a close ap-

proximation of a plug-flow reactor, which is the type of reac-

tor to be selected for reactions of a positive order.

2.

All

parts of the catalyst bed must contribute maximally

to the overall conversion, implying that all catalyst particles

must be in good contact with the reactant stream.

To be representative

of

an industrial reactor that comes

close to this ideal, the labo ratory reactor must also fulfill these

requirements. Although improper design may give rise to in-

dustrial reactors that p erform m ore poorly, it does not make

much sense to try to mimic a malfunctioning reactor on a

small scale. Therefore, compliance with the preceding re-

quirem ents for an ideal reactor can be used a s a guiding prin-

ciple in downscaling

of

fixed-bed reactors, and in the follow-

ing sections we will examine how far reactor dimensions can

be reduced without violating these requirements.

Limits in Downsizing

of

Fixed-bed Reactors

Deviations from

plug

f tow

The allowable devia-

tion fro m ideal plug flow is determined by what is considered

to be an acceptable difference in conversion rate achieved in

the ideal and the real reactor, and is also a function of the

ord er of the reaction and th e degree of conversion. How much

the residence time distribution in a practical reactor resem-

bles that in an ideal reactor is generally expressed by the

PCclet number (Pe ), a dimensionless number that is numeri-

cally about twice the equivalent number

of

mixers in series.

Fo r a true plug-flow reactor both num bers a re infinitely large.

Mea rs 1971) has derived a relation between Pe , the reac-

tion order

n ,

and the conversion X based on the criterion

that a good practical reactor should require no more than

5 extra catalyst to achieve the sam e conversion as the ideal

one. A similar relation has been proposed by Ge rm an 1988),

which is based on the somewhat less stringent criterion that

the te mp erature requirement for the sam e conversion should

not be higher than theoretical by one degree Centigrade,

which is considered to be t he accuracy

of

temperature defini-

tion in practice. The latter relation, which will be used

throughout this article, can be written as

Allowable D eviation

from

Plug Flow.

AIChE

Journal December 1996 Vol. 42 NO. 2 3499


3/10

in which

L

is the reactor length,

u

the superficial velocity,

and

D ,

the overall axial dispersion, which is made up from

contributions from axial molecular diffusion, convective dis-

persion in the packing, and macroscopic velocity distributions

in the reactor.

This can be an important cause

of spread in residence time in the reactor, particularly with

gases of high diffusivity and a t low linear velocities low space

velocities and shallow beds). For a rather demanding case,

namely, hydrogen at ambient conditions at a gas hourly space

velocity (GHSV) of 1,000, he minimum bed length needed to

satisfy the criterion of Eq. 1 for a good reactor is shown in

Figure

1 . It

follows that a microreactor with a bed length of

10cm will in most cases be acceptable.

The statistical variations

in

width, length, and direction of individual channels within the

packing result in a dispersion that is characterized by the di-

mensionless Bodenstein number Bo

=

d

*

u/D,, , hich is

analogous to the Ptc let number but has the particle diameter

d f as characteristic dimension.

For axial dispersion in the packing, Eq. 1 can be written as

Axial M olecular Diffusion.

Axial Convectiue Dispersion.

-

- ----

-

n = 2

n = I

-

-___

-

L / d f >

8/Bo)*

n

* In {1/(1- X I ) .

2)

For flow through packed beds

Bo

is a function of Re, and

a global correlation between these dimensionless numb ers has

been established by Gierman

on

the basis of published data

Gierman, 1988). According to this correlation Bo decreases

with decreasing

Re

until an approximately constant value of

about

0.4

is obtained in the range of interest for small labora-

tory reactors. For the axial dispersion of the liquid in trickle

flow, the constant value of Bo is an order of magnitude lower,

or about 0.04. It is of interest to note that for trickle-flow

processes as used in the petroleum industry the dispersion of

the gas phase is generally unimportant, since the gas phase

consists mostly of recycled hydrogen. Consequently, conver-

sion of hydrogen per pass is relatively low and rather unim-

portant, so that only the dispersion of the liquid-oil phase

needs to be considered.

L , c m

12

1

1

10

1

/I

t

85 90 95

100

CONVaKlON ,

o A i

I I 1 I I I

I

Figure

1.

Minimum bed length as a funct ion of conver-

sion for a first- and second-order reactor as

determined by molecular diffusion in flowing

gas.

Hydrogen at 1 bar and

25°C;

G H S V = 1,000 nL/ L.h); =

0.4; I = 2.

3500 December 1996

L.cm

(single phase

f low)

L,cm trickle flow)

5oE

1

85 90 95 100

CONVERSION

,

Figure 2. Minimum bed length as a function of conver-

sion for a first- and second-order reaction as

determined by axial d ispersion n the packing.

Figure 2 presents the minimum bed length calculated ac-

cording to Eq.

2

with the values for B o mentioned earlier, for

beds of particles of 2 mm diameter a typical size for fixed-bed

catalysts). It can be seen that in the case

of

single-phase flow,

for examp le, gas flow, a bed length

of

20 cm as can be accom-

modated in a microreactor is adequate for not too demand-

ing cases conversions not exceeding 90 ). However, for

trickle-flow operation much longer beds are required. For in-

stance, for

90

conversion in the case of a second-order re-

action the minimum b ed len gth of 2 m is required, w hich is in

the range of pilot-plant reactors. Hence, microreactors are

clearly unsuitable for testing catalysts of particle size in the

millimeter range in trickle flow, and pilot-plant reactors se em

indicated.

However, as can also be inferred from Figure 2, from the

point of view of axial dispersion there should be little prob-

lem in testing particles of 0.1 mm diameter in a 20-cm-long

bed in a m icroreactor even for trickle flow. This is one of the

motives behind the catalyst dilution technique to be dis-

cussed later.

Wall Effect and Other Causes of Transverse Macroscopic

Ve-

locity Profiles. In addition to the axial dispersion discussed

earlier, which is an intrinsic characteristic

of

the packing and

which finds its origin in the more or less random variation

of

velocity vectors on the scale of the particle diameter, more

systematic differences in velocities may occur in the flow

through the bed.

A

lateral velocity profile over the reactor

diam ete r may be caused by nonuniform packing uneven

compaction or segregation of particles in the case of a cata-

lyst with a spread in particle size).

Although the preceding lateral velocity profile may be

avoided by careful filling of the reactor, the re is another, more

inherent cause of systematic velocity differences, nam ely, the

wall effect. The packing

of

particles cannot be the same very

Vol. 42, No. 12

AIChE

Journal


4/10

close to the wall of the reactor as the unperturbed packing

farther away from it. Generally, there is a much greater

voidage in the wall region than the average voidage in the

interior of the reactor, as found both experimentally Bena ti

and Brosilow, 1982) and in computer-simulated packings

Zimmerman and Ng, 1986). Th e grea ter permeability in the

wall zone, which tends to enhance the flow in combination

with the flow retardation by the reactor wall, results in a ra-

dial velocity profile with a maximum located one to a few

particle diam eters away from the wall Fahien and Stankovic,

1979; Vortm eyer a nd Sch uster, 1983).

The wall effect gives rise to an appreciable spread in resi-

dence time of the reactant. However, as the ratio between

bed diameter and particle diameter increases, the flow

through

the

wall zone will carry less weight, and at a suffi-

ciently large ratio its effect becomes acceptably small. At a

ratio higher than about 25 the wall effect no longer has a

noticeable effect on the average properties of the bed, for

example, average porosity and average permeability Chu and

Ng, 1989).

Figure 3 shows the relation between the minimum reactor

diameter and the particle diameter with a value of

25

as a

criterion for an acceptably small wall effect. This figure can

also be read as the maximum particle diameter that can be

applied in a reactor of a given diameter. According to this

figure, a fixed-bed catalyst of a practical size in the millime-

ter rang e can only be tested in reactors of several centime ters

width if the effect of the wall is to be minimal. Hence, for

testing catalysts of such a practical size, a pilot-plant reactor

with a diameter of at least 4 cm seems indicated, and a mi-

croreactor with a diameter of

l

crn or less is clearly unsuit-

able. However, with respect to the wall effect there should be

no problem in such a microreactor with much smaller parti-

cles, for example,

0.1

mm diameter, and this is a further rea -

son

for applying the catalyst-bed dilution techn ique t o be dis-

cussed later.

dp,mm

0 1 2 3 4 5 6 7 8

Figure 3. Maximum part ic le size for a given b ed diam e-

ter, as dictated by the cri terion for n egl igible

wall effect.

0

m

Reducing the EfSect of Lateral Flow Profiles by Lateral Difsu

sion. Th e preced ing criterion for negligible wall effect is un-

duly stringent in the specific case that there is a sufficiently

fast exchange of mass between regions of low and high for-

ward velocities. This applies specifically to fixed-bed reactors

operated with gases, which have a relatively high molecular

diffusivity. Fast lateral mass exchange reduces the spread in

residence tim e caused by the lateral velocity profile, a nd the

interaction between axial velocity differences and lateral ex-

change results in a dispersion that can be described by an

axial diffusivity known as the Taylor diffusivity.

Th e length equivalent

to

one mixing stage associated with

the T aylor diffusivity can be w ritten as

in which u is the average interstitial velocity, R the bed ra-

dius, Dradhe diffusivity in the radial direction; is a dimen-

sionless number that characterizes the flow profile. Combin-

ing Eqs.

1

and

3,

one ob tains

in which E is the bed voidage.

The axial dispersion in packings caused by the combined

effects of lateral flow profiles and radial diffusion has been

the subject of many studies in chromatography. Information

obtained in these studies is relevant to the present problem

since the requirements of an ideal packed chromatographic

column, namely, minimal band broadening and adequate

contacting of the solid phase, are identical to those of an

ideal fixed-bed reactor. In studies with packed chromato-

graphic columns it has been found that for columns with a

low ratio of tube diam eter over particle diam eter D/dp <

5 )

the value of is abou t 0.04, which is abou t twice the value of

1/48 for the parabolic Poiseuille profile

of

laminar flow in a

tub e Sie and R ijnders, 1967). Using the value of 0.04 and

assuming the mechanism of radial diffusion to be molecular

diffusion through the bed, we have calculated the maximum

bed diameters for which radial diffusion is effective in wiping

out the effect of the wall. No te that the re is a maximum

limit on the bed diameter, whereas in the absence of fast

radial diffusion, there is a

minimum

diameter for a negligible

wall effect.) From results presented in Figure

4

t can be seen

that with gas as flowing medium there should be no problem

in testing catalyst particles of a practical size in a microreac-

tor of cm diameter, notwithstanding the fact that the diam-

eter ratio may be smaller than 5 .

On

the other hand, the

diffusivity in liquids is too low to rely o n lateral exchange for

effective counteraction of the wall effect in microreactors,

since the maximum bed diameter for this to happen is in the

millimeter range Figure 4).

Measurement of residence-time distribution in microreac-

tors ope rated with gas have confirmed that as a consequence

of fast radial diffusion the contribution of the wall effect to

the residence-time distribution is a relatively minor one. D ata

shown in Table

2

show that this contribution is negligible as

comp ared t o axial molecular diffusion.

AIChE

Journal December

1996

Vol.

42,

No.

12

3501


5/10

D.cm

Figure 4. Maximum b ed diameters for effect ive cou nter-

act ion of rad ial f low pro f i les by radial di ffu-

sion.

Beds w i th D / d p <

5.

Gas : Hydrogen,

P = 1

bar ;

T =

25°C;

G H S V

=

1,000 nL/ L.h) . Liquid: n-Eicosane. T

=

300°C;

WH SV weight hour ly space veloci ty)= kg/ kg.h). E =

0.4,

r = 2

Catalyst contacting

The second requirement of an ideal reactor, namely, that

all catalyst particles are in good contact with the reactant

stream, will in most cases

of

a single fluid phase be fulfilled if

the first criterion

of a sharp residence time distribution is

met. H owever, in the case of two flowing phases a sh arp resi-

dence-tim e distribution is

no

guara ntee for adequate contact-

ing. In a trickle-bed reactor a situation may prevail where

liquid flows preferentially through a certain part of the bed,

while gas flows mainly through the other part. Such macro-

scopic maldistribution of liquid can be observed in experi-

ments with colored liquids, but can also be simulated with

comp uter models Zimm erman a nd Ng, 1986). This maldistri-

bution, often referred to as incomplete wetting, may be sus-

tained if the force

of

gravity responsible for th e dow nflow of

liquid is much larger than the frictional forces. If the latter

forces predominate, the high pressure drop will force liquid

to flow through every available interstitial channel, and con-

sequently catalyst contacting will be good.

Table 2. Measured Axial Dispersion i n a Microflow Reactor

with Gas Flow*

No. of

Dax Dmo,/T

Packing Pe Mixers cm2/s)

(cm2/s>

Spheres 105 53 0.132 0.137

d , =

1.5

mm,

E = 0.40

d = 1.5 mm

L = 3 m m ,

c =

0.44

Cylinders 97 49 0.143 0.151

L

=

10.2

cm,

=

0.8

cm,

V

=

5

mL. Nitrogen,

P

=

1 bar ,

T

=

25T ,

G H S V

=

450 nL/ L.h) . Trace r : He. Binary dif fus ion coeff icient

of

H e in

N,

=

0.678 cmZ/s, r

=

2 assumed) .

A criterion for adequate wetting or even irrigation of cata-

lyst is the following one proposed by Gierman and Harmsen

Gierman, 1988)

in which

qL

is the dynamic viscosity of the liquid, pL the

liquid density, and g the gravity constant. The dimensionless

constant

W

is the wetting number that compares frictional

and gravitational forces. The numerical value in the preced-

ing equation has been determined on the basis of a large

volume of experimental data , and its order of magnitude can

be rationalized by simple considerations Sie, 1991).

Figure

5

shows the minimum bed length as a function of

particle diameter required to meet the wetting criterion just

given. The figure may also be read as the maximum particle

diameter that can be applied in a bed of a given length. It

can b e seen that, fo r liquid viscosities in the rang e typical for

petroleum fractions under hydroprocessing conditions, a min-

imum length of several meters is required with particles hav-

ing a dia me ter of 2 mm and larger. Hence, for testing of such

catalysts unde r trickle-flow conditions, a relatively large pilot

plant reactor is required a nd a m icroreactor can be ruled out.

However, the figure also shows that with much smaller parti-

cles, for example, 0.2-mm diameter and smaller, the required

bed length for good wetting of th e catalyst is less than

20

cm.

Hence, with such small particles microflow reactors become

eligible for testing catalyst for trickle-flow processes. This

constitutes a third argument for the catalyst dilution tech-

nique to be discussed below.

L

. m

/ I

rl

I I L I I

I

i l l 1 1

0.5 5

10

dp.mm

Figure 5. Minimum bed length for part ic les of varying

size, as dictated by cri terion for even irr iga-

tion.

LHS V l iquid hour ly space veloci ty)

= 1

L/ L.h).

3502

December 1996

Vol. 42, No. 12

AIChE Journal


6/10

Catalyst Bed Dilutionwith Fine Particles of

Inert

Material

From the foregoing discussion it transpired that for fixed-

bed processes operated with gas flow, reactors as small as

microreactors about I-cm diameter,

5

to 20-cm bed length)

can be used for testing catalysts of realistic size (1- to 3-mm

diameter). In contrast herewith, relatively large pilot-plant

reactors several meters high, several centime ters dia me ter)

are required for testing such catalysts for trickle-flow proc-

esses.

In principle, one has the option of crushing the industrial

catalyst and testing it in the form of fine particles in small

reactors. How ever, aside from the danger of affecting catalyst

properties by the crushing and sieving operations, the data

obtained with crushed catalyst may not be representative for

industrial practice. If pore-diffusion limitation occurs with the

practical catalyst, the crushed sample will exhibit a higher

activity. But even when the main reaction is not or is only

slightly limited by intraparticle diffusion, other reactions in

complex reaction networks involving a multitude of different

feed molecules may be more seriously limited by pore diffu-

sion. Hence, changing the catalyst particle size can affect not

only the apparent activity of the catalyst, but selectivity and

deactivation behavior as well.

The dilemma

of

wanting to test relatively large industrial

catalyst particles and needing fine particles in small beds for

the hydrodynamic reasons discussed earlier can be solved by

surrounding the catalyst particles with small granules

0.05-0.2-mm diameter) of an inert material. Thus, the hy-

drodynamics of the flowing fluids will be mainly dictated by

the packing of small inert particles, whereas the catalytic con-

1ISp-

l l S f

’*

I I W H S V ,

h

Figure 6. Second order plot for hydrodesulfurization of

a light gas

oil

in trickle flow over undiluted

beds of different lengths.

D =

8

cm. Catalys t : Co/Mo/Alumina, 3-mm ellets.

AXChE

Journal December 1996

version behavior is that of the catalyst in the actual size. The

improvement in plug-flow approac h

of

laboratory trickle-flow

reactors by adopting this catalyst-bed dilution technique has

been demonstrated in experiments with isotopically labeled

oils und er typical hydroprocessing conditions Van Klinken

and van Dongen,

1980).

Another beneficial effect

of

catalyst-

bed dilution is that temperature homogeneity in the bed is

improved and that the chances of temperature runaway are

diminished in th e case of strongly exothermic reactions, esp e-

cially when using inert diluents with high thermal conductiv-

ity, such as silicon carbide or alundum.

The technique of catalyst-bed dilution opens the way to

testing of industrial fixed-bed catalysts in very small labora-

tory reactors, even for trickle-flow processes. This is sup-

ported by th e results of bed-length reduction in gas-oil hydro-

treating experiments, shown in Figures

6

and

7.

From the

data for an

undiluted

bed shown in Figure 6, it can be in-

ferred that shortening of a pilot-plant reactor from 4.8-m

length by a factor of 3 is accompanied by a significant deteri-

oration of the apparent catalyst performance. This can be

understood from the previous discussion, since the length of

the shorter reactor

1.6

m) is below the minimal lengths dic-

tated by acceptable axial dispersion and by catalyst wetting

cf. Figures 2 and

5).

By contrast, reduction of the length of a

diluted

cataIyst bed by a factor

of

2, namely, from

40

cm to

20

cm, does not noticeably affect the test results, as can be in-

ferred from Figure

7.

In addition to supporting the earlier

thesis that extraparticle mass-transfer effects play a minor role

only, these results indicate t hat testing of diluted catalysts for

trickle-flow processes is feasible in microreactors, since the

5

Figure

7.

Second order plot for hydrodesulfurization of

a

gas

oil in trickle flow over diluted beds of

different lengths.

D

=

2 cm. Catalyst: Co/Mo /Alumina, 1.5-mm cylindrical

ex

t rudates . Di luent : 0.2 m m SIC.

Vol. 42, No. 12

3503


7/10

length of the shorter catalyst bed

is

in the range covered by

microreactors. Further evidence for this conclusion will be

presented below.

Table 4. Comparison

of

Microflow and Commercial Results

for

Trickle-FlowHydrodesulfurizationof Gas Oil*

Reac tor Microflow Commercial

Representat iveness

of

Catalyst Test ing in

Microf low Reactors

Fixed-bed processes with reactants in the gas phase

From the previous discussion on th e limiting dimensions

of

lab ora toy fixed-bed reactors it followed that for these proc-

esses microreactors with a bed diameter of about

1

cm and a

bed length between 5 and 20 cm should give results tha t com e

close to those of an ideal reactor, except for the most de-

manding cases conversions very close to 100 ). Such results

should therefore be representative for an industrial process,

assuming that the latter is carried out in a well-designed re-

actor.

Evidence to support this statem ent is presented in Table

3,

which compa res microflow test results o n light paraffin isom-

erization a process with reactants in the gas phase) with data

on the same catalyst and feedstock from a commercial unit.

It can be seen that there is a quite satisfactory agreement

between the two sets of data. It is

of

interest to note that the

catalyst in the form of extrudates of 1.5-mm dia me ter has not

been diluted.

Trickle-flow processes

The suitability of microflow reactors with diluted catalyst

beds for testing of industrial catalysts used in trickle-flow

processes such as hydroprocessing of petroleum fractions is

demonstrated by the data of Table 4,which is a direct com-

parison between microflow test results and commercial oper-

ating data

on

hydrodesulfurization of the same gas oil over

the same catalyst under comparable conditions. In the mi-

croflow test the catalyst bed of 1.2-mm extrudates was di-

luted with 0.05-mm particles of silicon carbide. T he close cor-

respondence between the two sets of data implies that the

data generated in the microreactor test can be considered to

be m eaningful for industrial practice.

Miniatur izat ionof Catalyst Test ing below th e Sc ale

of Microreactors

From the analysis of the limiting geometric factors, pre-

sented in the earlier part of this article, it can be concluded

that th e microreactor scale is not necessarily the sm allest scale

for representative testing of fixed-bed catalysts.

As

follows

Table

3.

Comparison

of

Microflow and Commercial Results

for Isomerization of Light Naphtha over a Ptmordenite

Catalyst (1.5-mm Extrudates)

Rea ctor Microflow Commercial

Catalyst charg e 2.1 g 15.9 t

Isopentane in product, 64.2 63.3

Isohexanes in produ ct, 81.9 84.8

2,2 diMe Butan e in 17.4 19.8

C , ,

yield, wt. 97.8 97.2

RON-0

of

produ ct calc.) 79.6 79.7

of total

C,

o f

total C,

product,

o f

total C,

Catalyst volum e

20 mL

diluted) 122 m3

Sulfur

in

produ ct, wt. 0.075 0.082

Desulfurization 95.4 95.0

Second-order rate constant

23.5

20.9

*Catalys t : Co/Mo/Alumina, 1.2-mm tr i lobe extrudates . D i luent : 0.05 rnm

Sic . Feed: Middle Eas t heavy gas o i l , 1 .64

wt.

S.

Operat ing condi -

t ions WH SV , WA BT weighted average bed tempera ture) , hydrogen/oil ,

par t ial pressures of hydrogen a nd hydroge n sulf ide at reactor inlet ) same

in both cases .

from the earlier analysis, it is in principle possible to reduce

the reactor diameter of about

1

cm typical for a microreac-

tor) down to a few millimeters, while keeping the bed length

about the same. The lower limit is in fact determined by the

requireme nt t hat th e catalyst particles have to be accom mo-

dated in the reactor, for example, for 1.5-mm-dia extrudates

the minimum inner diameter of the reactor may be about 2

mm. This implies a reduction in reactor and catalyst volume

by a factor

of

about

25,

that is, instead of a typical catalyst

volume of

5-10

mL in microflow reactors, the catalyst vol-

ume in the submicro- or “nanoflow” reactor am ounts to as

little as 0.2 to 0.4 mL.

Even on the latter small scale, meaningful test results can

be obtained, as will be demonstrated below.

Fixed-bed processes with reactants in the gas phase

Table

5

presents comparative test data obtained in mi-

croflow and nanoflow reactors for the isomerization of light

paraffins over a catalyst shaped as extrudates of 1.5-mm di-

ameter. The nanoflow test data match the microflow test data

quite well. The latter test data have been shown to be repre-

sentative for catalyst performance under industrial condi-

tions, and by implication this should also be true for the

nanoflow test results.

Figures 8 and 9 compare microflow and nanoflow test re-

sults on t he hydrodesulfurization of the same gas oil over a

catalyst shaped a s 1.5-mm extrudates, d iluted with 0.05-mm

particles of silicon carbide. It can b e seen tha t th ere is a good

match between the data obtained with the two types of reac-

tors, bo th with respect to desulfurization r ate Figure

8)

and

the effect of tem per atu re on desulfurization Figure 9).

Table

5.

Comparison of Microflow and Nanoflow Test

Results for Pen tan epex ane Isomerization*

Rea ctor Microflow Nanoflow

Catalyst charge , g 5 0.2

Isopentane

in

produ ct, 58.3 59.5

of

total pentanes

of total hexanes

product, of total hexanes

Isohexanes in product, 80.7 80.2

2,2 diMeButane

in

16.4 16.5

C 5+ yield, wt.

99.0 98.9

RON-0 of produ ct calc.) 77.8 77.9

*Catalys t : Pt /Mo rdeni te , 1.5-mm extrud ates undi luted) . Fe ed : n-pen -

tane/n-hexane/CycloC, 61/34/5 wt.

76 . T =

260°C; P

=

25 bar ; hydro-

gen/feed

=

1.25 molar ; WHSV

=

1.70.

3504 December 1996 Vol. 42, No. 12 AIChE Journal


8/10

/sxp- /s f

.8

2.6

2l4

2.2

4 1

-

-

-

-

-

-

-

-

OMICROFLOW

NANOFLOW

I

I

I I

0.1

0.2 a3

l /WHSV,

h

Figure

8.

Microf low vs. nanof low test resul ts for hydro -

desulfu rizat ion of gas oil in tr ickle f low; effect

of s pace velocity.

Catalyst: Co/Mo/Alumina, 1.5-mm cylindrical extrudates.

Di luent : 0.05

mm

S i c .

x = 0.84.

Since it has been shown that microflow test data can be

representative for industrial operation, it follows from the

preceding comparison that nanoflow test data, too, can be

meaningful for industrial practice. This statement is substan-

tiated by Table 6, where such test data on hydrodesulfuriza-

tion of a gas oil over a catalyst shaped as 1.5-mm extrudates

are directly compared with data from a commercial hydro-

desulfurization unit.

Potential errors in catalyst sampling

Proper sampling from a catalyst batch composed of non-

identical catalyst particles is obviously important for obtain-

ing meaningful results. To obtain representative samples

of

such an inhomogeneous catalyst batch for testing in labora-

tory reactors ranging from pilot plants to microreactors, the

well-known procedures for sampling

of

solid granular mate-

rial have to be adhered to. However, with a catalyst sample

as small as that to be tested in nanoflow reactors, an addi-

tional problem may arise due to the small number of parti-

cles.

A

catalyst sample of 0.3 g, consisting of 1.5-mm-dia. and

4-mm-long extrudates will typically consist of about

40

parti-

cles, and at this small number the law of statistics has to be

considered, Figure 10presents the calculated standard devia-

tions for a random sampling from a mixture of two catalysts

A and B ) that differ in activity by

a

factor k,/k,. Assuming

that a standard deviation below 5 is acceptable, it can be

seen that there is no problem when the sample size consists

AIChE Journal December 1996

l n

301

1.58

1.60 1.62

1.64

10001T.

K-1

Figure

9.

Microf lo w vs. nanof iow test results for hydro-

desulfu rizat ion of g as oi l in tr ickle f low; effect

of

temperature.

Catalys t and di luent as s t a t ed und er Figure 8.

of 10,000 particles, as long as the sampling is nonselective.

However, at a sample size corresponding with

20

particles,

the standard deviation may be above

5

when the activity

ratio is larger than a bout 1.5 and when the less active catalyst

constitutes a major proportion of the mixture. If the less ac-

tive catalyst is a minor component of the mix,

for

example,

lo ,

the standard deviation remains below the set limit of

5 , even at a very large activity ratio.

The latter situation may occur in practice because a batch

of catalyst can be contaminated with a small amount of less

active or even inert material. In such a situation, the limited

number of particles

in

the test sample does not necessarily

pose a problem, as long as the sampling is done in a nonse-

lective manner.

Table

6.

Comparison of Nanoflow Test Results with Data of

Commercial Operation on Hydrodesulfurizationof Gas Oil*

Rea ctor Nanoflow Commercial

Catalyst charge 0.228g + diluent) 33.5 t

Sulfur in product, wt. 0.155

0.144

Desulfurization

80.5 81.9

2.1-Order rate constant 22 24

*Catalys t : Co/Mo /Alumina, 1.5-mm t r i lobe extrudates . Di luent : 0.05 m m

S i c . F eed : G as o i l

162-415 C, 0.794

wt.

S.

Condit ions

WHSV,

WABT, hydrogen/oi l and par t ial pressures of hydrogen and hydrogen

sulf ide at reactor

inlet)

s am e

for

t he two cases.

Vol.

42, No. 12 3505


9/10

STANDARD DEVlATlON

IN 6

2l

1 1.2 1.4 1.6 1.8 2.o

Figure 10. Sampl ing errors f rom a catalys t mix ture

composed of two com ponents hav ing d i f fer-

ent activit ies, for samples cons is t ing of N

part ic les.

UI

kA kB

How ever, when ther e is a large activity difference and when

the less active material is abundant the statistical error may

be unacceptably large for small samples. Fortunately, this is

not a situation of much practical interest since such an inho-

mogeneous catalyst is obviously not an optimal one and there

is hardly a reason to try and assess the perform ance

of

such a

poor batch of catalyst with great accuracy.

Implementation and Consequences

of

Small-Scale

Catalyst Testing

The increased possibilities of catalyst testing on a smaller

scale tha n considered feasible in the past, and the advan tages

attached to experiments on a small scale, have led to a grad-

ual replacemen t of the traditional large pilot plants by smaller

units. This can be seen from Table 7, which compares the

distributions of reactor units of varying size in a n oil-process

R & D laboratory in 1970 and 1987. It can be seen that dur-

ing the intervening period bench-scale and microflow units

have taken over many of the tasks formerly reserved for larger

pilot plants. In more recent years, microflow tests are being

widely applied, not only in exploratory studies where in the

screening of new catalysts and unconventional feedstocks the

small sample sizes are very importan t, but also in the process

research and development studies on more established proc-

esses. In addition to the previously cited processes, namely,

paraffin isomerization and hydrotreating of distillates, mi-

Table

7.

Distribution

of

Number

of

Reactor Units*

Year 1970 1987

Microflow z

0.01

L) 13 50

Bench Scale

= 0.1

L)

20

38

Pilot Plant > 1 L) 67 12

Unattended

units

13 94

*At

the

oil process R &

D

Department of the Koninklijke/Shell Labora-

torium, Amsterdam in of total).

3506

December

1996

croflow reactors play an important role in research and de-

velopment on most hydrocarbon-conversion processes, in-

cluding catalytic reformin g Sie and Blau who ff, 19911, hydro-

cracking Van D ijk et al., 1991), and even hydrodemetalliza-

tion and hydroconversion of residual oils Oelde rik et al.,

1989). Th e use of m icroflow reactors is no t restricted to sim-

ple once-through experiments in isothermally operated reac-

tors, since on this scale accurate simulation

of

adiabatic oper-

ation h as proven feasible Sie and B lauwh off, 19911, as well

as simulation of a hydrocracker with on-line fractionation and

recycle of fractionator bottoms Van Dijk et al., 1991).

Driving forces behind this trend have been the improved

research a nd dev elopment efficiency resulting from savings

in

costs, materials, and personnel, as well as the enhancement

of safety and environm ental friendliness of the operation. T he

savings in materials in moving from pilot plant to microreac-

tor testing can be illustrated by comparing the feed require-

ments of these two reactor units for a typical case of gas-oil

hydrotreating: whereas the pilot plant consumes monthly

amounts of oil and hydrogen gas that have to be supplied on

a periodic basis by tank cars and tube trailers, the microflow

needs in a similar period can be covered by a jerry can of oil

and a few gas bottles. In th e pilot-plant case, the storage and

disposal of the large amount of unused products may also

constitute a logistic problem with environmental aspects.

The enhanced intrinsic safety resulting from miniaturiza-

tion may be illustrated by the m aximum hazards of a nanoflow

reactor. When working with flammable gases or liquids, flow

rates are

so

small that they can only sustain a flame com para-

ble in size to that of a small candle. Working with synthesis

gas, the small flow rate may give rise to an emission of toxic

carbon monoxide that is of the order

of

that from a burning

cigarette. The maximum release of compressed gas from the

reactor inventory upon mechanical failure would be less than

from a pu nctu red bicycle tire and if this gas is explosiv e, its

detonating power can be compared with that of a small fire-

cracker. Hence, many of the safety systems mandatory in the

case of larger reactors fire detectors, flammable a nd toxic-gas

detectors, special ventilation, etc.) can be simplified or even

dispensed with.

Th e higher intrinsic safety of smaller reactor units makes it

easier to automate these units and to allow them to operate

around the clock withput human attention. The increase in

automation, which has accompanied the size reduction of the

reactor units in the period 1970-1987, is clearly shown in

Table 7, where the percentage of unattended units is seen to

increase from 13 to 94.

The increase in research productivity resulting from cata-

lyst testing on a smaller scale is illustrated by Figure 11. It

can be inferred tha t the more m odern, autom ated microreac-

tor unit operating without human attention requires an ord er

of magnitude less manpower than the traditional large pilot

plant it has replaced.

Concluding Remarks

It has been shown that microflow and even smaller reac-

tors can be used not only in fundamental kinetic studies with

powdered catalysts, but also for testing practical fixed-bed

hydroprocessing catalysts under industrially relevant condi-

tions, that is, with catalysts in their actual shape and size,

Vol. 42, No. 12

AIChE Journal


10/10

MANHOUR PER REACTOR HOUR

21

M

ICROFLOW

automated 1

\

\

1960 I910 1980 1990

Moo

YEAR

Figure 11. Manpower needs of different reactor units as

a function of the approximate introduction

year for general use in oil process research.

with practical feedstocks, and at conditions

of

space velocity,

temperature, and pressure that are comparable to those in

practice, yielding data that can match those of commercial

operation. This applies not only to fixed beds operated with a

stream of gas, but equally to fixed beds operated with liquid

and gas as trickle-bed reactors, provided the catalyst bed is

diluted with fine inert material. This experimentally verified

conclusion

is

supported by a theoretical analysis

of

the fac-

tors that limit the downscaling possibility of fixed-bed reac-

tors.

These small reactors can now be used for generating cat-

alytic performance data that some 25 years ago had to be

done in large fixed-bed pilot plants, at only a fraction of the

costs, materials usage, and personnel need associated with

the operation of the large pilot plants. Hence, for many cat-

alytic process studies there is no nee d for the traditional large

pilot plant any more. Nevertheless, cases will remain where

experiments on a large scale cannot be avoided, for example,

when large quantities

of

a new product are needed for mar-

ket development in anticipation of commercial production. A

large pilot plant is also needed when hydrodynamic factors

play a crucial role in determining t he chem ical conversions of

a catalytic process or when m echanical/physical features, such

as fouling of the bed by particulate matter, are important.

Another reason for a large and complex integrated pilot plant

is when the process configuration is complex and involves

separation steps and recycles of streams that cannot be

miniaturized in a representative way.

For the evolution in catalyst testing systems, where minia-

turization is increasingly offering capabilities at lower costs

but cannot make large pilot plants totally redundant, an anal-

ogy

can be drawn with the evolution in computing: while mi-

crocomputers are finding an ever increasing role, it is still

necessary to make use of mainframe and even supercomput-

ers in special applications.

Acknowledgment

The author gratefully acknowledges the valuable contributions to

this work of his former coworkers at the Koninklijkefihell-Labora-

torium Amsterdam, where most of the work was carried out. He is

particularly indebted to Messrs J. Glezer and

A.

F. de Vries for their

role in the development of the nanoflow equipment and

in

the exe-

cution of the comparative tests.

Notation

Dmol

molecular diffusivity,

m2.s-’

or cm2.s-1

E

=activation energy, kJ.mol-’

k

= reaction-rate constant

P =pressure, bar

Sf

=sulfur content in feed, wt.

S p

=sulfur content in product, wt.

T =temperature, degrees Centigrade

V

=catalyst bed volume, m3 or L

=kinematic viscosity, m2.s-l

T = ortuosity factor

Literature

Cited

Benati, R. F., and C. B. Brosilow, “Void Fraction Distribution in

Beds of Spheres,” AZChE J. 8, 359 (1982).

Chu, C. F., and K. M. Ng, “Flow in Packed Tubes with a Small Tube

to Diameter Ratio,”AIChE J. 35,148

(1989).

Fahien, R.

W.,

and

I.

M. Stankovic,

“An

Equation for the Velocity

Profile in Packed Columns,”Chern. Eng. Sci., 34, 1350 (1979).

Gierman, H., “Design of Laboratory Hydrotreating Reactors. Scaling

Down

of

Trickle-Flow Reactors,”

Appl. Catal. ,

43,

277 (1988).

Mears,

D.

E., “The Role of Axial Dispersion in Trickle-Flow Labo-

ratory Reactors,” Chem. Eng. Sci., 26, 1361 (1971).

Oelderik, J. M.,

S.

T. Sie, and D. Bode, “Progress in the Upgrading

of Petroleum Residue,” Appl. CatuL,

47,

(1989).

Sie, S. T., “Scale Effects in Laboratory and Pilot-Plant Reactors for

Trickle-Flow Processes,”

Rev. Znst. Fr.

Pit . , 46,

SO1 (1991).

Sie,

S.

T., and G. W. A. Rijnders, “Band-Broadening in Packed

Chromatographic Columns,”

Anal.

Chim.

A c f a ,

38,

3 (1967).

Sie,

S.

T., and P. M. Blauwhoff, “Laboratory Equipment and Proce-

dures for Evaluation of Catalysts in Catalytic Reforming,” Cafal.

Today, 11, 103 (1991).

Van Dijk,

A., A.

F. de Vries, J.

A.

R.

van

Veen,

W . H. J.

Stork, and

P. M. M. Blauwhoff, “Evaluation of Hydrocracking Catalysts in

Recycle Tests,” Cafal. Today, 11, 129 (1991).

Van Klinken,

J.,

and R.

H. van

Dongen, “Catalyst Dilution for Im-

proved Performance of Laboratoly Trickle Flow Reactors,” Chem.

Eng. Sci., 35,59 (1980).

Vortmeyer, D., and

J

Schuster, “Evaluation of Steady Flow Profiles

in Rectangular and Circular Packed Beds by a Variational Method,”

Chem.

Eng.

Sci. 38, 1691 (1983).

Zimmerman, S. P., and K. M. Ng, “Liquid Distribution in Trickling

Flow Trickle-Bed Reactors,”

Chern. Eng. Sci., 41,

861 (1986).

Manuscript received

Mar.

19, 1996, and revision received

May

31, 1996.

AIChE Journal

December

1996 Vol.

42,

No. 12

3507

Miniaturization of Hvdrowocessina Catalvst Testing System -Sie-1996-AIChE_Journal

Documents

Transcript of Miniaturization of Hvdrowocessina Catalvst Testing System -Sie-1996-AIChE_Journal